1. Field of the Invention
The present invention relates to a fuel injection system. Specifically, the present invention relates to a fuel injection system for automatically correcting a deviation in an injection quantity, which is caused by a change with time and the like, by performing learning control.
2. Description of Related Art
In recent years, in accordance with reinforcement of exhaust gas limitation, higher and higher accuracy has been required in an injection quantity of fuel injection. For instance, in recent years, a diesel engine is required to perform a pilot injection, multi-step injection and the like because of the reinforcement of the exhaust gas limitation. Therefore, the accuracy in the injection quantity has to be improved.
In order to achieve the high injection accuracy, minute adjustment of a fuel injection device or the like can be performed before shipment. However, even if the minute adjustment is performed, there is a possibility that the injection quantity changes because of a change with time. In this case, there is a possibility that the high injection accuracy cannot be maintained.
Learning control is known as one of countermeasures against the above problem. In the learning control, a learning injection is performed during an operation of the engine, and a deviation between a learning injection quantity (an injection quantity calculated by a control device) and an actual injection quantity (an actually injected quantity) is calculated. Then, a learning value (a correction value) is calculated from the deviation and the injection quantity during a normal operation is corrected based on the learning value so that an injection quantity (an aimed injection quantity) calculated in accordance with an operating state of the engine coincides with the actual injection quantity.
For instance, in conventional learning control disclosed in Unexamined Japanese Patent Application Publication No. H10-205372, the learning injection (for instance, a small-amount injection) is performed if the operating state of the engine becomes a learning operating state. Then, multiple influence values representing the influence of the learning injection are obtained. The influence values are calculated from a change in the operating state of the engine (for instance, a change in rotation speed sensed by a rotation speed sensor). After the multiple influence values are obtained, a learning value for correcting the injection quantity during the normal operation is calculated based on an average of the obtained influence values.
In the conventional learning control, if the learning condition is established, the learning injection quantity (a learning master injection quantity) of the fuel is injected. The learning master injection quantity is an injection quantity suitable for the learning.
Therefore, if the learning master injection quantity of the fuel is injected in a state in which the injection accuracy is reduced because of the change with time and the like, there is a possibility that an engine operating state characteristic value (for instance, exhaust emission, engine noise or torque generated by the engine) becomes inappropriate.